The programme before
Northfield sells production-planning software to mid-market manufacturers. The target list was 340 accounts across the UK and DACH, split into three tiers, and the paid programme was roughly £22,000 a quarter, almost all of it on LinkedIn: a set of single-image and document ads pointed at a matched-audience upload of the list.
Reporting was a monthly export from Campaign Manager into a spreadsheet, joined to the target list with a lookup on company name. It said the programme was performing: CTR above the benchmark the agency had given them, a CPC that had been falling for two quarters, and a steady trickle of demo requests.
The problem was that none of those numbers were about accounts. They were about people, and the people they were about were not evenly distributed across the list.
What the first coverage report showed
The first thing Kairo produced was not a dashboard. It was a count: of 340 accounts on the list, 129 had never been delivered a single impression in the previous 90 days. Not low engagement — no delivery at all.
The reason turned out to be structural rather than mysterious. Three things were happening at once:
- The matched audience was smaller than the list. Around 11% of rows had failed to match at upload, mostly subsidiaries listed under a trading name rather than a registered one, and nobody had checked the match report after the first upload eight months earlier.
- The auction was doing the segmentation. One campaign covered all three tiers. LinkedIn optimised delivery towards the cheapest available impressions, which were concentrated in Tier 3 — the smaller companies, with less competition for their attention.
- Frequency capping was absent. A subset of accounts was being served fourteen or more times a month while a third of the list was on zero, and the average frequency figure in the monthly report concealed both halves of that.
“The uncomfortable finding was not that the ads were bad. It was that most of the list had never seen them. We had been optimising creative for an audience we were barely reaching.”
What changed
Nothing about the budget, and almost nothing about the creative. The changes were all structural, and there were four of them.
1. The list was rebuilt against the match report
Kairo flags unmatched rows with a reason. Twenty-nine of the 37 failures were resolved by using the registered domain rather than the trading name, and six were genuine duplicates of accounts already on the list under a parent. The list went from 340 nominal accounts to 331 real ones with a 96% match rate.
2. Tiers were split into separate campaigns
The single biggest change. Once each tier had its own campaign and its own budget floor, the auction could no longer quietly reallocate Tier 1 budget towards cheaper Tier 3 impressions. Tier 1 CPMs went up by roughly 40%, as expected, and Tier 1 coverage went from 51% to 92% inside six weeks.
3. Frequency was capped, and the floor was raised
A cap of eight impressions per person per month on Tiers 2 and 3 freed enough delivery to cover the previously unserved accounts. This is the change that most teams resist, because the accounts you are capping are the ones that look most engaged in the old report.
4. Reporting moved to cost per account reached
The monthly review stopped opening with CTR. It opened with coverage by tier and cost per account reached by tier — two numbers that make an under-served segment impossible to miss.
Where it stands a quarter later
Unreached accounts are at 6%, and the remaining ones are mostly genuinely hard: three companies with almost no LinkedIn presence, and a handful where the buying committee sits in a subsidiary Northfield has not mapped yet. Tier 1 coverage has held above 90% for two consecutive quarters.
The number the board sees is cost per account moved into In market, reported alongside influenced pipeline on a stated 90-day engaged-touch definition. Both are lower-precision than the CTR figure they replaced, and both are about something the business cares about.
What they would do differently
- Check the match report every month, not at setup. Lists rot. Companies rename, get acquired, and change domains, and an eight-month-old match rate is fiction.
- Split tiers from day one. The single campaign was chosen for simplicity and it cost them two quarters of Tier 1 coverage.
- Agree the review metrics before the change, not after. The efficiency metrics getting worse was predictable, and having agreed in advance that it would happen is what stopped the programme being reverted.